We buy, price, and structure the behavioral data that frontier labs can't get anywhere else. That takes tooling to move petabytes of it and experiments to say what it's worth. We're hiring engineers for both.
The work
- Data processing tooling. Custom pipelines over petabyte-scale corpora across image, video, audio, documents, and code.
- Large scale batch inference for annotation and filtering, at a volume where cost per token is a design constraint.
- ML experiments that measure what a dataset actually does to a model downstream. That number is the one this whole market is missing.
What it takes
Most of this has no playbook yet. The job is turning a vague question about data quality into an experiment that answers it, then building the thing that runs it at scale.
A sign you'll like it here: you're open to dangerous ideas and doing hard things.
Reach out
Tell us what you've been working on.